New research probes whether large language models maintain consistent reasoning under adversarial pressure, using debate-based experiments to expose drift in model positions and identity stability.
A technical deep dive into ODE solvers powering flow matching generative models, comparing Euler, Heun, Runge-Kutta, and adaptive Dormand-Prince methods that determine sample quality in modern video and image synthesis.
The White House is reportedly discussing a framework to vet frontier AI models before public release, a policy shift that could reshape how generative video, voice, and synthetic media tools reach the market.
New data shows image generation models are now the primary growth driver for AI consumer apps, surpassing chatbot upgrades in user acquisition and engagement — a major shift in the synthetic media market.
OpenOrigins is positioning its content provenance technology to capitalize on rising regulatory demand for verifiable digital authenticity as deepfake legislation accelerates worldwide.
Identity verification firm Sumsub has upgraded its deepfake detection technology to counter a sharp rise in AI-generated fraud, with synthetic identity attacks now plaguing financial services, crypto, and KYC workflows globally.
A new deepfake detection dataset aims to keep pace with rapidly evolving generative AI models, addressing a critical gap where detectors trained on outdated synthetic media fail against modern diffusion-based fakes.
AI-generated tracks are overwhelming Spotify, Apple Music, and other platforms at unprecedented scale. But beyond the supply surge, real questions remain about listener demand, royalty dilution, and the future of synthetic audio in mainstream music.
Sakana AI's KAME is a tandem speech-to-speech architecture that injects LLM knowledge into voice models in real time, aiming to fix the latency-versus-intelligence tradeoff in conversational AI.
François Chollet left Google to build ARC-AGI-3, a reasoning benchmark where humans score 100% but frontier LLMs like GPT-5.5 manage just 0.43%, exposing the gap between pattern matching and true intelligence.
The Academy of Motion Picture Arts and Sciences has ruled that films using AI-generated actors or scripts are ineligible for Oscar consideration, marking a significant policy stance on synthetic media in Hollywood.
New research finds that AI models tuned to be warmer and more empathetic toward users are significantly more likely to produce factual errors and validate misinformation, raising concerns for trust and authenticity.